Visit the CSIO website at: www.csio.econ.northwestern.edu.E-mail us at: csio@northwestern.edu.

THE CENTER FOR THE STUDYOF INDUSTRIAL ORGANIZATION

AT NORTHWESTERN UNIVERSITY

Working Paper #0010

Commercialization of the Internet:The Interaction of Public Policy and Private

Choicesor

Why Introducing the Market Worked so Well

By

Shane Greenstein*

KGSM Department of Management and StrategyAnd NBER

April, 2000

* This paper cites liberally from research supported by the Council on Library Resources, the Kellogg GraduateSchool and the National Science Foundation. The author thanks Angelique Augereau, Oded Bizan, Tim Bresnahan,Barbara Dooley, Chris Forman, Amy Friedlander, Avi Goldfarb, Brian Kahin, Mort Rahimi, Scott Stern, MitchellWaldrop and many seminar participants for comments. I am particularly grateful to Zvi Griliches who encouragedthis research when it was at a formative stage. All remaining errors are mine alone.

Abstract

Why did commercialization of the Internet go so well? This paper examines events in the Internet

access market as a window on this broad question. The study emphasizes four themes. First,

commercializing Internet access did not give rise to many of the anticipated technical and operational

challenges. Entrepreneurs quickly learned that the Internet access business was commercially feasible.

Second, Internet access was malleable as a technology and as an economic unit. Third, privatization fostered

attempts to adapt the technology in new uses, new locations, new market settings, new applications and in

conjunction with other lines of business. These went beyond what anyone would have forecast by examining

the uses for the technology prior to 1992. Fourth, and not trivially, the NSF was lucky in one specific sense.

The Internet access industry commercialized at a propitious moment, at the same time as the growth of an

enormous new technological opportunity, the World Wide Web. As it turned out, the Web thrived under

market-oriented, decentralized and independent decision making. The paper draws lessons for policies

governing the commercialization of other government managed technologies and for the Internet access

market moving forward.

1

1. Motivation

The "commercialization of the Internet” is shorthand for three nearly simultaneous events: the

removal of restrictions by the National Science Foundation (NSF) over use of the Internet for commercial

purposes, the browser wars initiated by the founding of Netscape, and the rapid entry of tens of thousands of

firms into commercial ventures using technologies which employ the suite of TCP/IP standards. These events

culminated years of work at NSF to transfer the Internet into commercial hands from its exclusive use for

research activity in government funded laboratories and universities.

Sufficient time has passed to begin to evaluate how the market performed after commercialization.

Such an evaluation is worth doing. Actual events have surpassed the forecasts of the most optimistic

managers at NSF. Was this due to mere good fortune or something systematic whose lessons illuminate the

market today? Other government managed technologies usually face vexing technical and commercial

challenges that prevent the technology from diffusing quickly, if at all. Can we draw lessons from this episode

for the commercialization of other government managed technologies?

In that spirit, this paper examines the Internet access market and one set of actors, Internet Service

Provides (ISPs). ISPs provide Internet access for most of the households and business users in the country

(NTIA, 1999), usually for a fee or, more recently, in exchange for advertising. Depending on the user

facilities, whether it is a business or personal residence, access can involve dial-up to a local number or 1-800

number at different speeds, or direct access to the user's server employing one of several high-speed access

technologies. The largest ISP in the United States today is America-On-Line, to which approximately half the

households in the US subscribe. There also are many national ISPs with recognizable names, such as AT&T

Worldnet, MCI WorldCom/UUNet, Mindspring/Earthlink, and PSINet, as well as thousands of smaller

regional ISPs.

The Internet access market is a good case to examine. Facilities for similar activity existed prior to

commercialization, but there was reason to expect a problematic migration into commercial use. This activity

2

appeared to possess idiosyncratic technical features and uneconomic operational procedures which made it

unsuitable in other settings. The Internet’s exclusive use by academics and researchers fostered cautious

predictions that unanticipated problems would abound and commercial demand might not materialize.

In sharp contrast to cautious expectations, however, the ISP market displayed three extraordinary

features. For one, this market grew rapidly, attracting thousands of entrants and many users, quickly

achieving mass-market status. Second, firms offering this service became nearly geographically pervasive, a

diffusion pattern rarely found in new infrastructure markets. And third, firms did not settle on a standard

menu of services to offer, indicative of new commercial opportunities and also a lack of consensus about the

optimal business model for this opportunity. Aside from defying expectations, all three traits -- rapid growth,

geographic pervasiveness and the absence of settlement -- do not inherently go together in most markets. The

presence of restructuring should have interfered with rapid growth and geographic expansion. So explaining

this market experience is also interesting in its own right.

What happened to make commercialization go so well? This paper's examination reveals four themes.

First, commercialization did not give rise to many of the anticipated technical and operational challenges.

Entrepreneurs quickly learned that the Internet access business was commercially feasible. This happened for

a variety of economic reasons. ISPs began offering commercial service after making only incremental

changes to familiar operating procedures borrowed from the academic setting. It was technically easy to

collect revenue at what used to be the gateway functions of academic modem pools. Moreover, the academic

model of Internet access migrated into commercial operation without any additional new equipment suppliers.

Second, Internet access was malleable as a technology and as an economic unit. This is because the

foundation for Internet inter-connectivity, TCP/IP, is not a single invention, diffusing across time and space

without changing form. Instead, it is embedded in equipment which uses a suite of communication

technologies, protocols and standards for networking between computers. This technology obtains economic

value in combination with complementary invention, investment and equipment. While commercialization did

3

give rise to restructuring of Internet access to suit commercial users, the restructuring did not stand in the

way of diffusion, nor interfere with the initial growth of demand.

Third, privatizing Internet access fostered customizing Internet access technology to a wide variety

of locations, circumstances and users. As it turned out, the predominant business model was feasible at small

scale and, thus, at low levels of demand. This meant that the technology was commercially viable at low

densities of population, whether or not it was part of a national branded service or a local geographically

concentrated service. Thus, privatization transferred the operation of the technology to a new set of decisions

makers who had new ideas about what could be done with it. Since experimentation was not costly, this

enabled attempts to adapt the technology in new uses, new locations, new market settings, new applications

and in conjunction with other lines of business. While many of these attempts failed, a large number of them

also succeeded. These successes went well beyond what anyone would have forecast by examining the

limited uses for the technology by non-commercial users prior to 1992.

Fourth, and not trivially, the NSF was lucky in a particular sense of the word. It enabled the

commercialization of the Internet access industry at a propitious moment, at the same time as the growth of

an enormous new technological opportunity, the World Wide Web. This invention motivated further

experimentation to take advantage of the new opportunity, which, as it turned out, thrived under market-

oriented and decentralized decision making.

The paper first develop these themes. Then it describes recent experience. It ends by discussing how

these themes continue to resonate today.

2. Challenges during technology transfer: an overview

Conventional approaches to technological development led most observers in 1992 to be cautious

about the commercialization of the Internet. To understand how this prediction went awry, it is important to

understand its foundations.

4

Many studies of the commercialization of technology emphasize the situated nature of technological

development. Technologies do not simply spring out of the ether; instead, learning processes and adaptation

behavior shape them. Users and suppliers routinely tailor technologies to short term needs, making decisions

that reflect temporary price schedule or idiosyncratic preferences, resulting in technological outcomes that

can only be understood in terms of these unique circumstances and origins.1 Such themes resonate

throughout studies of technologies which develop under government management.2

1 The literature on general purpose technologies (Bresnahan and Trajtenberg [1995], Helpman [1998]) also

helps frame these themes by highlighting the role of co-invention, defined as the complementary inventions whichmake advances in general purpose technologies valuable for particular organizations in particular places at particularpoints in time.

2 For example, see studies of the supersonic transport (Cohen and Noll [1990]), nuclear power (Cowan[1988]), air frames (Mowery and Rosenberg [1992]) and the early history of computing (Flamm [1989], Goody-Katz andPhillips [1982]), among many such examples.

Seen through this light, the most problematic feature of the Internet was its long exclusive use by

military, government or academic users. Prior to 1992 it had developed into the operations found at an

academic modem pool or research center. These were small scale operations, typically serving no more than

several hundred users, involving a mix of frontier and routine hardware and software. A small operation

5

required a server to monitor traffic and act as a gate-keeper, a router to direct traffic between the Internet and

users at PCs within a local-area-network (LAN) or calling center, and a connection to the Internet backbone

or data exchange point operated by the NSF. These were often run by a small staff, either students or

information technology professionals.

Revenues were not regularly collected in these arrangements and budgetary constraints were not

representative of what might arise with commercial operations and competitive pressures. Many small

colleges had opened their Internet connections with NSF subsidies. The organizational arrangement within

research computing centers also was idiosyncratic, usually with only loose ties, if any, to the professionally

run administrative computing centers of a university or research organization. The array of services matched

the needs of academic or research computing, which had only a partial overlap with the needs of commercial

users.

Any student of technology transfer would have confidently predicted that the transition into

commercial markets would give rise to challenges. Standing in 1992 and looking forward, it was uncertain

whether these challenges would take a long time to solve and whether commercial user’s needs would be

difficult to address. In general, conventional analysis anticipates one of three challenges: technical,

commercial and structural challenges.

Technical challenges often arise during commercialization. Government users, government

procurement and government subsidies result in technology with many features mismatched to commercial

needs. Products possessed features for which vendors or users have no need. Alternatively, commercial

vendors and users do need other features. Thus, as a technical or engineering matter, a technology which is

mature for exclusive non-commercial uses -- such as a military application -- may appear primitive in civilian

use. It may require complementary inventions to become commercially viable. If these requirements are

considerable, then commercialization may occur slowly.

For example, military users frequently require electronic components to meet specifications that suit

6

the component to battle conditions. Extensive technical progress is needed to tailor a product design to meet

these requirements. Yet, and this is difficult to anticipate prior to commercialization, an additional amount of

invention is often needed to bring such a product design and to bring its manufacturing to a price/point with

features that meet more cost-conscious or less technically stringent commercial requirements.

Commercial challenges arise when commercial markets require substantial adaptation of operation

and business processes in order to put technologies into use. In other words, government users or users in a

research environment often tolerate operational processes that do not translate profitably to commercial

environments. After a technology transfers out of government sponsorship, it may not be clear how to

balance costs and revenues for technologies that had developed under settings with substantial subsidies

underwriting losses, and research goals justifying expenditures. Hence, many government managed

technologies require considerable experimentation with business models before they begin to grow, if they

grow at all.

For example, the supersonic transport actually met its engineering targets, but still failed to satisfy

basic operational economics in most settings. Being technically sleek was insufficient to attract enough

interest to generate the revenue which covered operating costs on any but a small set of routes. No amount of

operational innovations and marketing campaigns were able to overcome these commercial problems.

New technologies are also vulnerable to structural challenges that impede pathways to

commercialization. Commercial and structural challenges are not necessarily distinct, though the latter are

typically more complex. Structural challenges are those which require change to the bundle of services

offered, change to the boundary of the firms offering or using the new technology, or dramatic change to the

operational structure of the service organization. These challenges arise because technologies developed

under government auspices may presume implementation at a particular scale or with a set of technical

standards, but require a different set of organizational arrangements to support commercial applications.

For example, while many organizations provided the technical advances necessary for scientific

7

computing in academic settings during the 1950s, very few of these same firms migrated into supporting

large customer bases among business users. As it turned out, the required changes were too dramatic for

many companies to make. The structure of the support and sales organization were very different, and so too

were the product designs. Of course, the few who successfully made the transition to commercial users,

such as IBM, did quite well, but doing so required overcoming considerable obstacles.

In summary, conventional analysis forecasts that migrating Internet access into commercial use

would engender technical, commercial and structural challenges. Why did the migration proceed so different

than expected?

3. The absence of challenge in the Internet Access industry

An ISP is a commercial firm who provides access, maintains it for a fee and develop related

applications as users require. While sometimes this is all they do, with business users they often do much

more. Sometimes ISPs do simple things such as filtering. Sometimes it involves managing and designing email

accounts, data-bases and web pages. Some ISPs label this activity consulting and charge for it separately;

others do not consider it distinct from the normal operation of the Internet access services.

On the surface the record of achievement for ISPs is quite remarkable. Most recent surveys show

that no more than 10 percent of US households get their Internet access from university-sponsored Internet

access providers, the predominant provider of such access prior to commercialization. Today almost all users

go to a commercial providers (Clemente [1998], Ebring and Nie [2000]). As of 1997, this ISP industry was

somewhere between a three and five billion dollar industry (Maloff [1997]), and it is projected to be much

larger in a few years.

By the end of the century the ISP market had obtained a remarkable structure. One firm, America

On-line, provided access to close to half the households in the US market, while several score of other ISPs

provided access to millions of households and businesses on a nationwide basis. Thousands of ISPs also

8

provided access for limited geographic areas, such as one city or region. Such small ISPs accounted for

roughly a quarter of household use and another fraction of business use.

3.1 Technical challenges did not get in the way

The Internet access market did suffer from some technical challenges, but not enough to prevent

rapid diffusion. Commercialization induced considerable technical innovation in complementary inventive

activities. Much of this innovative activity became associated with developing new applications for existing

users and new users.

It is often forgotten that when the electronic commerce first developed based on TCP/IP standards, it

was relatively mature in some applications, such as e-mail and file transfers, which were the most popular

applications (these programs continue to be the most popular today, NTIA [1999]). To be sure, TCP/IP based

programs were weak in others areas, such as commercial data base and software applications for business

use, but those uses did not necessarily have to come immediately. The invention of the World Wide Web in

the early 1990s further stretched the possibilities for potential applications and highlighted these weaknesses.

More important for the initial diffusion, little technical invention was required for commercial vendors

to put this technology into initial mainstream use. Academic modem pools and computing centers tended to

use technologies similar to their civilian counterparts -- such as bulletin board operators -- while buying most

equipment from commercial suppliers. Moving this activity into the mainstream commercial sector did not

necessitate building a whole new Internet equipment industry; it was already there, supplying goods and

services to the universities and to home PC users. Similarly, much of the software continued to be useful –

i.e., Unix systems, the gate-keeping software, and the basic communication protocols. Indeed, every version

of Unix software had been TPC/IP compatible for many years due to Department of Defense requirements.

A simple commercial operation only needed to add a billing component to the gate-keeping software to turn an

academic modem pool into a rudimentary commercial operation.

9

Technical information about these operations was easy to obtain if one had sufficient technical

background; a BA in basic electrical engineering or computer science was far more than adequate. Many ISP

entrepreneurs had used the technology as students or in related lines of business. Descriptions of some of the

earliest access operations show that they did not employ any exotic hardware or rare technologies (Kalakota

and Whinston [1996], Kolstad [1998]). Many Internet bulletin boards quickly developed and Boardwatch

Magazine, among others, expanded its focus from bulletin boards to ISPs as early as 1994, also spreading

information about how to operate such ventures. Several vendor associations, such as the Commercial

Internet Exchange, were formed and also served as information sources.

Users with investments in networking technology, such as LANs or simple client/server

architectures, also could adopt basic features with little further invention. Internet technologies associated

with textual information had incubated for twenty years and were well past the necessary degree of technical

maturity necessary for mainstream use. Telnet, FTP and the basic protocols for email were widely diffused

and relatively easy to use. Some communication software already used TCP/IP and many of the common

programs could easily adapt to it. There were already many similar technical activities taking place in

commercial settings. TCP/IP compatibility was built into Windows 95, which further eased investments for

users after 1995.

The basic commercial transaction for Internet access also did not raise prohibitive technical issues.

Most often it involved repetitious and on-going transactions between vendor and user. A singular transaction

arose when the vendor performed one activity, setting up Internet access or attaching Internet access to an

existing computing network. If the ISP also operated the access for the user, then this on-going operation

provided frequent contact between the user and vendor, and it provided frequent opportunity for the vendor

to change the delivery of services in response to changes in technology and changes in user needs. This

worked well because in many cases an ISP was better educated about the technological capabilities than the

user. In effect, the ISP sold that general knowledge to the user in some form that customized it to the

10

particular needs and requirements of the user. At its simplest level, this provided users with their first

exposure to a new technological possibility while educating them about its potential.

Often access went beyond exposure to the Internet, especially with a business user, and included the

installation, maintenance and training, as well as application development. These types of transfers of

knowledge typically involved a great deal of nuance, often escaped attention, and yet, were essential to

developing infrastructure markets as an on-going and valuable economic activity. The basic technical know-

how did not differ greatly from routine knowledge found in the computing services sector prior to

commercialization.

Finally, some NSF decisions and legacy regulatory decisions also aided. When the NSF took over

stewardship of the Internet backbone, it invested in developing a scalable system of address tables and IP-

address systems. Subsequent growth tested those investments and inventions; no surprising problems were

found, nor did any engineering problems hinder growth. Domain name registration also remained a gentle

monopoly until recently. Data exchange points remained organized around the cooperative engineering

principles used within the NSF days. A competitive data communications industry was beginning to reach

adolescence at about the same time as commercialization and provided additional access points for new firms,

particularly in urban areas. So as a technical matter, interconnection with the public switch network did not

pose any significant engineering challenges (Werbach [1997]).

3.2 Commercial challenges did not slow diffusion

Internet access was built in an extremely decentralized market environment. Aside from the loosely

coordinated use of a few de facto standards (such as the World Wide Web consortium), government

mandates after commercialization were fairly minimal. ISPs had little guidance or restrictions. They were

therefore able to tailor their offerings to local market conditions and to follow entrepreneurial hunches about

growing demand.

11

As a technical matter, there were few barriers to entry in the provision of dial-up access. As a result,

commercial factors, and not the distribution of technical knowledge among providers, largely determined the

patterns of development of the basic dial-up access market immediately after commercialization. To the

surprise of many, the operational procedures developed over two decades lent themselves to the early

commercial implementations, fostering a foundation for commercial growth. As with many new markets

which spawn in non-commercial environments (Ventresca, et al [1998]), many features were borrowed

wholesale and without question. In effect, entrepreneurs borrowed the organization of the academic modem

pool and tried to put a revenue generating function on top it. Billing software was added to the basic gateway

component, and once this proved to be a feasible way to collected revenue, many entrepreneurs built on top

of that commercial form.

Shortly after commercialization in 1994, only a few commercial enterprises offered national dial-up

networks with Internet access, mostly targeting the major urban areas. Pricing was not standardized and

varied widely (Boardwatch, 1994-1995). Most of these ISPs were devoted to recreating the type of network

found in academic settings or modifying a commercial bulletin board with the addition of backbone

connections, so interconnection among these firms did not raise insoluble contracting or governance

problems. These ISPs were devoted primarily to dial-up; few ISPs attempted sophisticated data-transport

over higher-speed lines, where the regulatory issues could be more complex and where competitive local

exchange competitors were developing the nascent market.

Very quickly ISPs learned that low cost delivery required locating access facilities close to

customers. This had to do with telephony pricing policies across the US. The U.S. telephone system has one

pervasive feature; distance-sensitive pricing at the local level. In virtually every part of the country, phone

calls over significant distances (i.e., more than thirty miles) engender per-minute expenses, but local calls are

usually free. Hence, Internet access providers had a strong interest in reducing expenses to users by providing

local coverage. Unmet local demand was a commercial opportunity for an entrepreneurial ISP.

12

As it turned out, access over dial-up lent itself to small scale commercial implementations. Several

hundred customers could generate enough revenue to support physical facilities and a high-speed backbone

connection in one location, so scale economies were not very binding. The marginal costs of providing dial-up

services were low and the marginal costs of expansion also fell quickly, as remote monitoring technology

made it cheap to open remote facilities. The marginal costs to users of dial-up service were also low in

response, involving only incremental changes for organizations that had experience with PC use or LAN

technology. It was easy to generate revenue in subscription models, where a commercial firm withheld

availability of access unless payment was made. Hence, the economic thresholds for commercial dial-up

service turned out to be feasible on a very small scale, encouraging small firms and independent ISPs. To be

sure, many firms also tried to implement access businesses on a large scale, but the economic advantage of

large scale did not preclude the entry of small scale firms, at least not at first.

Finally, decades of debate in telephony had already clarified many regulatory rules for interconnection

with the public switch network, eliminating some potential local delays in implementing this technology on a

small scale. The FCC treated ISPs as an enhanced service, not passing on access charges to them as if they

were competitive telephone companies, effectively making it cheaper and administratively easier to be an ISP.

This decision did not receive much notice at the time since most insiders did not anticipate the extent of the

growth that would arise. As ISPs have grown and as they threaten to become competitive voice carriers,

these interconnection regulations have come under more scrutiny (Sidek and Spulber [1999], Weinberg

[1999]).3

3 If anything, regulatory decisions for reciprocal compensation of competitive location exchange providers

(CLECs) encouraged CLEC entry, which also partly encouraged ISP entry through interconnection with CLECs.

13

In retrospect, two key events of 1995 set the stage for the commercial ISP market for the remainder

of the decade. The first was the Netscape IPO in August 1995. The other was the entry of AT&T World Net.

Though important to incumbent local exchange carriers, however, one should not exaggerate this too much. Thescale of this phenomenon grew tremendously in the late 1990s, but ISP entry started well before then. Moreover,since CLEC entry was primarily concentrated in dense urban areas, much of this effect was felt in urban areas, whichwould have experienced a great deal of ISP entry even without this implicit subsidy to CLECs.

The World Wide Web was known in the academic community in the early 1990s. It began to diffuse

prior to commercialization and accelerated with Mosaic, an prototype browser developed at the University of

Illinois. Many ISPs included Mosaic on their systems. Despite licensing the technology to many firms, the

University of Illinois did not generate as much excitement as the Netscape IPO, which brought extensive

publicity to the new technology (Cusumano and Yoffie, 1998). The subsequent browser wars further

heightened this awareness.

The emergence of the web changed the commercial opportunities for ISPs. ISPs found themselves

both providing a "traditional" service in demand, text-based applications such as e-mail, and trying to position

themselves for a new service, web applications. This new opportunity provided strong incentives to grow and

experiment with new business models and new lines of service. It also induced considerable new entry. While

not all markets experienced the same type of competitive choices, nor did all ISPs see the same opportunities,

many private firms found ways to develop opportunities quickly, seeding lessons which they then applied in

other localities.

AT&T's entry was also important but its actions mattered because of what did not happen rather than

14

what did. AT&T developed a nationwide Internet access service, which was available in much of the

country, opening with as large a geographic spread as any other contemporary national provider. It also grew

quickly, acquiring one million customers with heavy publicity and marketing. This growth depended on the

strength of its promise to be reliable, competitively priced and easy to use. It was deliberately aimed at

households, and provided a mass-market service from a name brand. It was a commercial success, to be

sure, but that was all. It was not a huge or dominant success, nor did it initiate a shake-out or restructuring of

the market for ISP service.

Here was a branded, nation-wide, professionally operated subscription model of ISP service, opening

with as large a geographic spread as any other contemporary national provider. Yet, it did not end the growth

of others, such as AOL, nor did it stop new entry of small firms, such as Mindspring, nor did it initiate a

trend towards consolidation around a few national branded ISP services. In other words, even with its deep

pockets AT&T did not dominate the offerings from all other firms, nor did it end the restructuring of the

access business. This defied many predictions about how this market would be structured, further

encouraging the decentralized growth and the emergence of independent ISPs.

Growth and entry brought about extraordinary results. Downes and Greenstein [1998] have

constructed maps that illustrate the density of location of ISPs at the county level for the fall of 1996 and

1998; Black and white versions of these are at the back of this paper.4 For color versions see, respectively:

http://www.kellogg.nwu.edu/faculty/greenstein/images/htm/Research/Maps/mapsep1.pdf,

http://www.kellogg.nwu.edu/faculty/greenstein/images/htm/Research/Maps/mapoct98.pdf.

4 For further documentation of these methods, see Downes and Greenstein [1999] or Greenstein [1999]. The

fall 1996 data covers over 14,000 phone numbers for over 3,200 ISPs. The fall 1998 data cover over 65,000 phonenumbers for just under 6,000 ISPs.

15

Colored areas are counties with providers. White areas have none. As the maps show, ISPs tend to

locate in all the major population centers, but there are also plenty of providers in rural areas. The maps also

illustrate the importance of changes over time. Many of the areas that had no coverage in the fall of 1996

were covered by the fall of 1998. Many of the areas that had competitive access markets in the early period

were extraordinarily competitive in the latter period.

Downes and Greenstein [1998] show that more than 92 percent of the US population had access by a

short local phone call to seven or more ISPs by 1998. No more than five percent did not have any access.

Almost certainly the true percentage of the population without access to a competitive dial-up market is much

lower than five percent. In other words, with the notable exception of some low-density areas, ISP service

was quickly available everywhere. To put it simply, among the vast majority of the US population in urban

and suburban areas lack of use was primarily due to demand factors, not the absence of supply.

An unexpected pattern accompanied this rapid growth in geographic coverage. First, the number of

firms maintaining national and regional networks increased over the two years. In 1996, most of the national

firms were recognizable; they were such firms as IBM, AT&T, and other established firms who entered the

ISP business as a secondary part of their existing services, such as providing data services to large corporate

clients. AOL, CompuServe and Prodigy all were in the process of converting their on-line service, previously

run more like bulletin boards than ISPs, into Internet providers. By 1998, many entrepreneurial firms

maintained national networks and few of these new firms were recognizable to anyone other than an industry

expert.

There was also a clear dichotomy for growth paths of entrepreneurial firms who became national and

regional firms. National firms grow geographically by starting with major cities across the country and then

progressively moving to cities of smaller populations. Firms with a regional focus grow into geographically

16

contiguous areas, seemingly irrespective of urban or rural features.5

Most of the coverage in rural areas comes from local firms. In 1996, the providers in rural counties

with under 50,000 population were overwhelmingly local or regional. Only for populations of 50,000 or

above, do national firms begin to appear. In fall of 1998, the equivalent figures were 30,000 or lower,

indicating that some national firms had moved into slightly smaller areas and less dense geographic locations.

In other words, Internet access in small rural towns is largely done by a local or regional providers, with

national firms only slowly expanding into similar territory.

5 Some ISPs have told me in interviews that this growth was initially in response to customer requests for

local phone numbers for accessing networks (e-mail mostly at first) when these customers traveled outside theirprimary area. More recently, it is also common to have ISPs discuss the possibility of developing a large customerbase for purposes of "selling the base" to a high bidder in some future industry consolidation.

It appears as if it does not pay for many large national providers to provide dial-up service for the

rural areas whereas many small local firms in other lines of business (e.g., local PC retailing) can afford to

add Internet access to their existing business. It may also be the case that the local firm may have an easier

time customizing the Internet access business to the unique needs of a set of users in a rural setting.

3.3. What structural challenges arose?

Commercialization of the Internet created an economic and business opportunity for providing

access. The costs of entry into low quality dial-up access were low, and commercially-oriented firms filled

voids in specific places. For any firm with national ambitions, coverage of the top fifty to one hundred cities

17

in the US was a fleeting advantage and quickly become a necessity for doing business. For any local or

regional firm in an urban market, many competitors arose.

Yet, not long after the Netscape IPO the ISP industry began to enter a second phase. Profitability and

survival involved more than geographic expansion. It involved bringing ISP service to the households and

businesses with PCs, but without access. Related, it also involved expanding into services which took

advantage of new opportunities associated with the web.

Understanding this second phase requires an understanding of the services ISPs offer other than

basic access and how those began to evolve. These new services include one of several activities: monitoring

technical developments, distilling new information into components which are meaningful to unfamiliar users,

and matching unique user needs to one of many new possible solutions enabled by advancing technical

frontiers. Sometimes it includes heavy use of the technological frontier and sometimes not. In general, it

depends on the users, their circumstances, their background, their capital investments, the costs of adjusting

to new services, and other factors which influence the match between user needs and technological

possibilities.

ISPs commercialized their adaptive role by offering new services, which can be grouped into five

broad categories: networking, hosting, web page design, basic access and frontier access (see appendix of

Greenstein [1999], for precise definitions.).

Networking involves activities associated with enabling Internet technology at a users' location. All

ISPs do a minimal amount of this as part of their basic service in establishing connectivity. However, an

extensive array of these services, such as regular maintenance, assessment of facilities, emergency repair, and

so on, are often essential to keeping and retaining business customers. Note, as well, that some of these

services could have been in existence prior to the diffusion of Internet access.

Hosting is typically geared toward a business customer, especially those establishing virtual retailing

sites. This requires the ISP to store and maintain information for its access customers on the ISP's servers.

18

All ISPs do a minimal amount of hosting as part of basic service, even for residential customers (e.g., for

email). However, some ISPs differentiate themselves by providing an extensive array of hosting services,

including credit-card processing, site-analysis tools, and so on.

Web design may be geared toward either the home or business user. Again, many ISPs offer some

passive assistance or help pages on web page design and access. However, some offer additional extensive

consulting services, design custom sites for their users, and provide services associated with design tools and

web development programs. Most charge fees for the additional services.

Basic access constitutes any service as slow as or slower than a T-1 line. Many of the technologies

inherited from the pre-commercial days became standard parts of basic access and were not regarded as a

new service. A number of other new functions, such as audio streaming, filtering and linking, also gradually

became standard parts of most firm's offering. Frontier access includes any access faster than a T-1 line,

which is becoming the norm for business access. It also includes ISPs that offer direct access for resale to

other ISPs or data-carriers and ISPs that offer parts of their own "backbone" for resale to others.6

By 1998, different ISPs had chosen different approaches, offering distinct combinations of services

and distinct geographic scopes. Table 1 shows the results of a survey of the business lines of 3816 Internet

service providers in the United States who advertise on thelist, an on-line directory of ISPs, in the summer of

1998 (see Appendix of Greenstein [1999]). Virtually every firm in the sample provides some amount of

dial-up or direct access and basic functionality, such as e-mail accounts, shell accounts, IP addresses, new

links, FTP and Telnet capabilities, but these 3,816 seem to under-represent both very small and quasi-public

ISPs (e.g., rural telephone companies).

6 Speed is the sole dimension for differentiating between frontier and basic access. This is a practical choice.

There are a number of other access technologies just now becoming viable which are slow but technically difficult,such as wireless access. Only a small number of firms in this data offer these services and these firms also offer highspeed access.

19

Table 1Product lines of ISPs

Category definition Most common phrases in category OriginalSample

Providing and servicingaccess though differentchannels

28.8, 56k, isdn, web TV, wireless access, T1, T3, DSL, frame relay, e-mail,domain registration, new groups, real audio, ftp, quake server, IRC, chat, videoconferencing, cybersitter TM.

3,816 (100%)

Networking Service andMaintenance

Networking, intranet development, WAN, co-location server, network design,LAN equipment, network support, network service, disaster recovery, backup,database services, novell netware, SQL server

789 (20.6%)

Web Site Hosting Web hosting, secure hosting, commercial site hosting, virtual ftp server,personal web space, web statistics, BBS access, catalog hosting

792 (20.7%)

Web Page Developmentand Servicing

Web consulting, active server, web design, java, perl, vrml, front page, secureserver, firewalls, web business solutions, cybercash, shopping cart, Internetmarketing, online marketing, electronic billing, database integration

1,385 (36.3%)

High Speed Access T3, DSL, xDSL, OC3, OC12, Access rate > 1056k 1,059 (27.8%)

Of the 3,816 ISPs, 2,295 (60.1%) have at least one line of business other than basic dial-up or direct

Internet access. Table 1 shows that 1,059 provide high speed access,789 networking, 792 web hosting, and

1,385 web page design. There is some overlap: 1,869 do at least one of either networking, hosting or web

design; 984 do only one of these three; 105 do all three as well as frontier access. This reveals many different

ways to combine non-access services with the access business.7

3.4 The contours of response to structural challenges

Structural issues did not become resolved quickly and have not disappeared as of this writing. This

occurred because these activities contain much more complexity and nuance than Table 1can display.

ISPs customize Internet technologies to the unique needs of users and their organizations, solving

problems as they arise, and tailoring general solutions to idiosyncratic circumstances and their particular

7 One of the most difficult phrases to classify was general "consulting." The vast majority of consulting

activity is accounted for by the present classification methods as one of these three complementary activities,networking, hosting and web-design.

20

commercial strengths. Sometimes ISPs call this activity consulting, and charge for it separately; sometimes it

is included as a normal business practice. In either case, it involves the translation of general knowledge

about Internet technologies into specific applications which yield economic benefits to end-users.

What factors influenced vendors' attempts to construct viable and on-going economic entities using

new technology in an evolving market place? Is it possible to classify and analyze the determinants of co-

invention? Why did some regions play host to ISP growth and others did not? There are many explanations,

but these aggregate into two classes, one which emphasizes firm-specific factors and another which

emphasizes location-specific factors:

Firm-specific factors: Firm-specific factors shape the incentives to bring new technology into use

(See, e.g., Demsetz [1988] or Nelson and Winter [1977] for a summary). ISPs came to the new

opportunities with different skills, experiences or commercial focus. In the face of considerable firm-specific

commercial uncertainty, ISPs purchased and installed their own capital equipment, publicized brand and

service agreements, and made other long-lasting investments. Many of these investments could commit the

ISP to particular services, even before market demand was realized or new commercial opportunities were

recognized.

Strategies pursued by national firms can be viewed in this light. Most national ISPs covered the same

geographic territories, so their strategies reflected either unique assets at the firm level, a firm's vision for

where their service should fall relative to competitors, or some other firm-specific feature. A more detailed

look at each of IBM, AT&T, AOL, Earthlink/Mindspring and PSINet will illustrate the variety of strategies

each pursued.

IBM had been an early entrant in the ISP market, focusing primarily on business customers and

secondarily on home users. Their service grew rapidly nationwide and globally, complimenting their

considerable other computer services. Yet, in a few years the firm decided to divest itself of its ISP backbone

and facilities, eventually selling to the highest bidder, AT&T. The firm concluded that joint provision of

21

access and other computer services was not a strategic advantage, and therefore focused its attention on

computer operations in many firms. The full benefits from this refocusing will only be manifest in time.

AT&T entered into consideration in another way. As already noted, it added a dial-up service soon

into commercialization. In 1998 it purchased TCI/@home, a cable company and Excite, a web portal. These

acquisitions position them for providing data service to the home with some content. With the recent

agreement to purchase Media One, which was pending at the FCC as of this writing, AT&T became the

largest cable provider in the country. The benefits from this are somewhat speculative, as the revenue stream

justifying these purchases has not been realized. If voice telephony, streaming media or any other host of new

broadband services become viable over cable lines, AT&T is well positioned to provide them. Subscription

fees for high speed access could also justify these purchases, if that technology becomes widely adopted.

AOL took a different approach. First, it grew its home user base through aggressive marketing to

more technical users. In response to the proliferation of ISPs in the mid 1990s, it ended its tiered subscription

model and introduced a flat-rate pricing model which mimics these other ISPs. Next it bought CompuServe, a

failed competitor with a loyal customer base, and currently operates it as a separate branded entity. It also

sold off its access facilities to Uunet, a sub-division of MCI/Worldcom, announcing concentration on the

development of content. It has since pursued its "walled-garden" strategy of making AOL proprietary content

attractive and the primary focus of AOL users. The purchase of ICQ, an instant messaging service, and

Time/Warner, among others, are consistent with this strategic approach. It is still an ISP, but a unique one,

providing access to the Internet that its customer base infrequently uses. The full benefits of this approach are

speculative as of this writing, as the revenues from it have not been fully realized.

Earthlink and Mindspring illustrate the issues facing new entrants on a national level. They market a

low-cost reliable service which is also easy to use, successfully competing against AT&T with much the

same appeal but a different branding. These firms also specialize in making the Internet easy to use for the

non-AOL user, the web-surfer who wants some but not too much help. Eventually these firms merged, partly

22

to consolidate their resources for competition against AOL, and partly to compete more strongly in the non-

AOL customer space. As one of the largest dial-up services in the country, there is a big question whether

they can survive in their niche in the face of competitive substitutes from all sides.

Finally, PSINet illustrates the feasibility of embarking on a strategy of emphasizing infrastructure.

They started as a consumer Internet service, but got out of that business in 1996. They had built out their

own backbone, investing in high speed facilities across the country, focused on becoming a carrier's carrier

for other ISPs and for business. Part of their strategy involves heavy investments in complimentary services,

such as hosting services corporate software services, which can offer high speed service when located just

next to fast Internet backbone lines. They also focus on offering infrastructure services to business,

developing services such as VPNs, which take advantage of their technical capabilities and nationwide

coverage. Once again, the full benefits of this approach are speculative, depending on realizing demand in the

future.

There are, of course, many other national firms. As with the above examples, their strategies mix

different elements of speculative investment, restructuring of organizations and entrepreneurial guesses about

future demand. In all cases, these experiments involve executives making investments under technical and

commercial uncertainty, restructuring production and distribution assets on a grade scale, trying to bring new

services to market, and only finding out if they meet market demand years after those investments.

It is also important to recognize the variance associated with local and regional ISPs, another and

particularly interesting subset of ISPs, who provide service for approximately between a fifth and a quarter

Internet users in the US. These firms locate in many different parts of the country; hence their firm-specific

strategies are also influenced by factors associated with their locations.

Location specific factors: A well-known line of economic research, dating at least to Griliches

[1957], has emphasized the geographic dispersion of incentives to adopt new technology. In this instance,

while basic dial-up access is widely available in all urban areas and many rural areas (Downes and Greenstein

23

[1998]), there is great variance in market structure on a local level. Some areas contain many suppliers from a

wide variety of backgrounds, while others contain few suppliers. From the standpoint of an ISP, many of

these structural features of markets are exogenous, and shape the competitive pressures on the ISP. In

addition, ISPs customize frontier technology to the needs of enterprises doing business at a specific time in a

specific place. The costs of this may vary by region because infrastructure differs by region. The demand

for higher speed service should also differ across regions if the users who find speed valuable are unevenly

distributed across geographic regions -- e.g., someone from San Francisco may be more willing to pay for

speed than people from Poughkeepsie.

The contrast between firm-specific and location-specific questions are examined in Augereau and

Greenstein [1999], who looked at small ISP’s investments in upgrades, and Greenstein [1999], who examined

small ISP’s propensity to offer services other than routine service associated with basic access. Both studies

identify the importance of geographic factors by taking advantage of variation between the locations of small

ISPs.

These studies are motivated by two observations. First, as noted in Downes and Greenstein [1998],

most large firms located in the same (or largely overlapping) set of major cities. Hence, for the importance of

location to be understood, the cause of variation between the small firms needs to be identified. Second,

Greenstein [1999] and Strover et al [1999] document that ISPs in rural locations tend to provide fewer high

quality services than those found in urban locations. Was this due to differences in infrastructure between

urban and rural areas, differences in the type of customer found there, or differences in the type of

entrepreneur who locate in different regions?

These studies both found that firm size, capacity and financial strength were important determinants

of behavior. There was also some evidence in Augereau and Greenstein [1999] that local infrastructure quality

influenced investment behavior. Generally, variation in local demographic conditions or competitive conditions

did not influence behavior. Both studies find much unmeasured variance in behavior, consistent with the

24

presence of unmeasured location-specific or firm-specific determinants. Moreover, the factors which lead

ISPs to offer new services, such as size, previous investments and strategic focus, are disproportionately

found in national firms and in local firms in urban areas.

These findings are consistent with the view that the scale of investment, the local infrastructure's

quality, and the explicit costs shape investment decisions by young ISPs in emerging markets. It is also

consistent with the view that there is too much commercial uncertainty in this market for firms to tailor the

technical vintages of their capital stocks too closely to geographically local demand or competitive conditions.

Finally, it is consistent with the view that most young firms with ambitious expansion plans initially locate in

urban areas instead of rural areas, growing their base markets and expanding outward, if at all.

4. Past lessons and future challenges

As public discussion of electronic commerce has grown, a loose coalition of prophets for the new

economy has come to dominate popular discussion. They write for such publications as The Industry

Standard, Business 2.0, Wired, Red Herring, Fast Company, and more "Web-zines" than anyone can list. It is

only a slight exaggeration to say that all popular portrayals of the Internet contain two principal features. First,

the prophets declare a business revolution in all information-intensive activities -- such as broadcasting,

entertainment, retail marketing, supply-chain management, other coordinative activity and research. Second,

and this is related, these same prophets proclaim that this technology's novelty dilutes standard lessons from

the past. In other words, because this technology contains so many unique features, it is ushering a new

commercial era which operates according to new rules.

To be sure, there is probably a grain of truth to these declarations. However, momentary euphoria

does not, nor should it, justify too simplistic a retrospective view of what actually happened, nor what is

about to happen. Indeed, this paper showed that a traditional economic perspective does provide considerable

insight into this new industry. In that spirit, we return to the questions which motivated the study and recap

25

the findings:

4.1 The commercialization of Internet access technology

• Why did the Internet access business grow quickly? Stated simply, exclusive use did not lead to

isolated technical and operational developments. Hence, commercializing Internet access did not give rise to

any difficult or insolvable technical and operational challenges. This was due in no small part to the way in

which the defense department and the NSF incubated the technology. It grew among researchers and

academics without being isolated from commercial suppliers. That is, the technology grew without generating

a set of suppliers whose sole business activity involved the supply of uniquely designed goods for military or

government users. Related to this was the fact that the basic needs of researchers and academics were not so

different from early commercial users. Hence, simple applications of the Internet invented for academic users

-- such as e-mail and file transfer using phone lines -- migrated to commercial uses without much technical

modification.

• Why did geographic ubiquity arise? To summarize, the Internet access business was

commercially feasible at a small scale and, thus, at low levels of demand. This meant that the technology was

commercially viable at low densities of population, whether or not it was part of a national branded service or

a local geographically concentrated service. Again, this partly mimicked the academic experience, where the

operations were also feasible on a small scale, but that statement alone does not capture all the factors at

work. Internet access was feasible in a wide variety of organizational forms, either large and small. Small

scale business opportunities thrive with the help of entrepreneurial initiative, which tend to be widespread

throughout the US -- including many low density and isolated cities in otherwise rural areas, which were

largely not being served by national firms. Small scale implementation also depended on the presence of high

quality complementary local infrastructure, such as digital telephony, and interconnection to existing

communications infrastructure. These too were available throughout most of the US due to national and local

26

initiatives to keep the communications infrastructure modern.

• Why did the Internet access business not settle into a common pattern? Market forces did

not impose uniformity in the use nor in the supply of access technology. Part of this was due to the absence

of technical and commercial challenges, which allowed low cost experimentation of the technology in new

uses, new locations, new market settings, new applications and in conjunction with other lines of business.

More generally, the technology was quite malleable as an economic unit. It could stand alone or become part

of a wider and integrated set of functions under one organizational umbrella. Such malleability motivated

experiments with new organizational forms for the delivery of access services, experiments which continue

today. Finally, and unique to this example, the invention of the World Wide Web brought new promise to the

technology. Not only did new business models arise to explore and develop its primitive capabilities and

expand them into new uses, but it motivated firms to experiment with Internet access along side new business

lines.

• Why did market forces lead to such extensive growth? This case illustrates how market forces

can customize new technologies to users and implement new ways of delivering technologies. These activities

have immense social value when there is uncertainty about technical opportunities and complex issues

associated with implementation. In addition, as the literature on general purpose technology would put it, co-

invention problems are best situated with those who face them. In this case, those actors were ISPs who

knew about the unique features of the user, the location or the application. More generally, commercialization

transferred development into an arena where decentralized and unregulated decision-making took over. This

was precisely what was needed to customize Internet access technology to a wide variety of locations,

circumstances and users. Removing the Internet from the exclusive domain of NSF administrators and

employees at research computing centers brought in a large number of potential users and suppliers, all

pursuing their own vision and apply it to unique circumstances. In addition, it allowed private firms to try new

business models, employing primitive web technologies in ways that nobody at the NSF could have imagined.

27

• In what sense did the NSF get lucky? As it turned out, the NSF commercialized the Internet

access industry at a propitious moment, during the growth of an enormous new technological opportunity,

the World Wide Web. Competitive forces sorted through new uses of this opportunity in particular places,

enabling some businesses to grow and unsentimentally allowing unsuccessful implementations to fade. To be

sure, some of these developments were heavily shaped by non-profit institutions, such as the World Wide

Web Consortium or the Engineering Internet Task Force, but profit motives still played a prominent role. Said

another way, had NSF stewardship over the Internet continued there would have been some experimentation

at computing centers found at universities and government laboratories, but it would not have been possible to

replicate all the exploratory activity that did arise in commercial markets.

4.2. Disentangling the systematic from the merely fortunate

While it was correct to forecast that commercial firms would restructure Internet access to suit

commercial users, many users did not need such restructuring to make use of the technology. Internet

access obtained widespread commercial appeal without restructuring of operations and other facets of supply.

As noted, this occurred for many reasons, but two historically unique factors heavily shaped the story. For

one, the Internet was a demonstrably viable network prior to its commercialization, already used by many

researchers, a fact that aided its migration into commercial use through incremental change. For two, the

invention of the web fueled commercial growth above and beyond what probably would have happened in any

event. Will broad lessons emerge in spite of these particular circumstances?

Said another way, while it is better to be lucky than right, it is always better to be both right and

lucky. Would NSF have been right if they were unlucky? What if the Browser had not been invented? Would

we still lauding the NSF for pursuing policies friendly to commercialization? In that spirit, this section briefly

considers two counter-factual questions: 1. Would outcomes have been similar in the absence of the browser?

2. Would outcomes have been similar in the presence of the browser, but the absence of NSF policies friendly

28

to commercialization?

• The importance of the browser: To answer a counter-factual question, it is important to ask

"compared with what alternative set of events?" This is difficult to answer in this instance because actual

events had a certain inevitability to them. For example, consistent with its mandate as a public research

institution, the University of Illinois encouraged diffusion of the browser through licensing (Cusumano and

Yoffie [1998]). To be sure, the Netscape browser of 1995 was a match thrown into a dry field, but parts of

that field had already been set ablaze. After University of Illinois began licensing Mosaic it was only a matter

of time before the blaze became an inferno. In other words, if Netscape had not commercialized the

technology somebody else would have done so soon. As another example, if Tim Berners-Lee had abandoned

his project before completion, it appears that somebody else eventually would have invented something

similar. Tim Berners-Lee's invention of hypertext (and then the world wide web) culminated decades of work

associated with making computing easier to use, more networked and more visual instead of textual (Waldrop

[2001]).

Hence, the browser and hyper-text appear to have a certain inevitability to them. In that light, the

most conservative counter-factual is this: What if hypertext and the browser had been invented a few years

later? Would the Internet have commercialized successfully?

The answer would appear to be yes, though events might not have been as dramatic. There are

several reasons for that asessment. First, email alone would have motivated considerable household adoption

of Internet access even without the browser. Email was among the most popular uses for the Internet in its

early years, so too were many of the community bulletin boards, financial applications, news, and chat

rooms. There were some substitutes for these activities even without the Internet, but email (especially)

would have been difficult to recreate in private networks on a national level and would have compelled some

commercial activity. Both households and business found this application useful and all surveys of Internet

use place it as the most popular application (Clemente, 1998, Nei and Lutz, 2000). While some of the more

29

visual applications in the bulletin board industry, such as commercial pornography and probably much

electronic retailing, would not have moved to the Internet without the browser, surveys such as Clemente,

1998, have never shown these as anything more than a fraction of early Internet use.

On a business level it is also possible to imagine considerable demand for Internet access even in the

absence of the browser. Many of the same applications just discussed, such as e-mail and news, motivated

business demand. In addition, much of the on-line data base industry would have found benefit from moving

to TCP/IP based file transfer as a substitute for bulletin board-based file transfers, which were more

cumbersome for users that a standard FTP or Telnet download. With some challenges to overcome,

commercial transactions which were forced into EDI-based data transfers also would have found TCP/IP

technology useful. However, it would have taken considerable time to shift many other data-base applications

into this mode, so one should not underestimate the difficulties (which were considerable even with the

browser). So it is reasonable to expect growth of TCP/IP connections within private industry even without

the browser, but not as fast.

Even with a later invention of the browser, many of the other institutions supporting the development

of the Internet also would still have been in place The creation of the Internet Engineering Task Force would

have continued to have an impact on standards development and diffusion. There might not have been

anything similar to the World Wide Web Consortium, but the shareware movement would have continued, a

factor that made it easier to obtain software for setting up independent ISPs. Related, the computing industry

had become sufficiently vertically disintegrated by the early 1990s to prevent any single firm from blockading

diffusion of TCP/IP8; neither IBM's proprietary networking offerings, nor DEC's, nor anyone else's could

8 The direction of commercial events also would have continued to take the same directions. Important

among them was the final dissolution of the working relationship between IBM and Microsoft, as well as the finaltriumph of Ethernet-based standards within the majority of networking equipment for LANs.

30

have dominate networking communications standards the way TCP/IP did once it began to commercialize.

Finally, even without the browser, one would have expected some migration of on-line capabilities

into commercial use at some level. Migration would not have been unusual by historical standards. New

computing capabilities often incubate among technically sophisticated users, building up functionality over

long periods of time before migrating into mainstream use (Bresnahan and Greenstein [1999]). In this

instance, the situation was ripe for migration. All the prototypes for text-based on-line activity existed among

sophisticated users. Moreover, the new functionality associated with Internet technologies did not require

radical investments on the part of users to be commercially viable. To be sure, there was one historical

novelty to the pattern of migration in this instance. Due to NSF restrictions on use, the sophisticated users of

Internet access technology were primarily concentrated in research positions and at universities, a subset of

sophisticated users in the computing industry. Aside from this feature, the broad pattern of incubation and

migration resembles other episodes of platform and technological growth in computing.

This is not to take away credit from those who took the actions and made them happen, nor to de-

emphasize the importance of these events for firms, regions and individuals. The contours of events most

certainly would have played out differently if the browser had diffused later. It would have resulted in very

different outcomes for particular companies, stock-holders and, arguably, regions where these companies

locate. Without the browser subscription models of Internet access might have had lower adoption rates at

business and homes, growth might not have been as explosive, and a different structure of supply might have

arisen. However, it is important to recognize the broad pattern which arises irrespective of the contours of

how it plays out: even without the browser Internet technology would have migrated into commercial markets

and demand would arise under any scenario, motivating the industry to continue to grow to a substantial level.

• The importance of NSF policies. Government employees deliberately let the baby bird out of the

nest, encouraging its flight. NSF's policies enabled the entrepreneurial initiatives of commercial firms to

influence migration of the technology. That said, migration of technology out of the research community into

31

mainstream commercial markets might have happened under many government policies. So the question

arises: Which government policies were critical? In light of later market events, the facet of NSF activities to

highlight are those policies which did not turn exclusive use of the Internet into an idiosyncratic technology

during its incubation.

There were many senses in which Internet technology was not isolated during its incubation. For

example, after the NSF created the NSFNET in the mid 1980s there were no attempts to exclude researchers

who had only mild research justifications for using the Internet, a policy decision which dated back to

conflicts which arose when DARPA managed the precursor to the Internet. The diffusion of TCP/IP in the

late 1980s further facilitated those goals, as it was an easy standard to use in virtually any computing

network. The NSF also did not isolate the Internet from mainstream computing use or vendor supply, making

contracts with firms such as IBM and MCI for operations, effectively subsidizing computing facilities at

research facilities which did the same. In addition, as the NSF developed and subsidized growth of the

Internet at many locations, adopting a decentralized set of "regional networks" for its operation. This structure

later facilitated private financing of Internet operations and further decentralization of investment decisions by

organizations with commercial orientation.

It is possible to view other events in late 1980s and early 1990s in a similar light. For one, NSF

contracted with third parties, such as MERIT, for operations. These types of contracts prevented the

network technology from being distant from mainstream engineering and technical standards. Related, NSF

permitted interconnection with private data communication firms, such as UUNet and PSI, a spin-off from

one of the regional networks, well before commercial dial-up ISPs came into existence. These contracts also

established precedents. Finally, NSF did not tightly police the use restriction, especially in the regional

networks. Indeed, a number of staff worked towards a 1992 congressional law which officially lifted the use

policy on NSFNET, providing more certainty that commerce could be conducted using assets which might

have appeared (to a court) to be previously owned by the federal government.

32

Finally, it is important to note the source of a particularly common error in large infrastructure

development, the attitude of "build it and they will come." That is, researchers and developers operating under

government subsidies tend to fulfill their own vision of what to do with the technology instead of a user's.

The NSF's actions effectively prevented this attitude from overwhelming development. As it turned out here,

the immediate use of Internet technology within academic research centers tended to put things to use

quickly. It allowed researchers find out what worked and why. Hence, some user desires influenced system

design, operation and growth -- even prior to the emergence of organizations who have a commercial

orientation and a direct incentive to take account for those desires. 9

9 One might ask why NSF adopted these policies when they did and whether their consequences were

anticipated. Unfortunately, that is a longer story and beyond the scope of this paper, which was simply to highlightthose which were useful in light of later events. The account of Waldrop [2001] , for example, begins that evaluationby arguing that NSF was making virtue out of necessity. He argues that there was no expectation that governmentagencies could operate a large scale data network indefinitely. This was particularly so at NSF, whose budget wasperiodically realigned by the whims and fads of political fashion. There also was no expectation that NSF could orwould fund decades worth of large scale data communications research on the scale that DARPA had done. Hence, itwas believed that a sustainable network would necessarily require private partnership on some level and, eventually,private financing.

33

34

In summary, commercialization of government managed technology can fail because there is no

incentive to anticipate technical, commercial or structure challenges which may arise later in commercial

markets. Since this failure can happen for a variety of reasons, it is not possible to point to any single one of

NSF actions as the policy which prevented such a failure or, alternatively, acted as the catalyst for

commercialization. It is, however, accurate to say that the sum total of NSF's actions did not let exclusive use

by researchers impose an irreversible idiosyncratic stamp on the Internet during formative periods of

incubation. These policies did not generate an isolated technology nor foster creation of a non-malleable

operation around it. Instead, NSF incubated technology with features that could adapt to the demands of

users who would later be in the majority. This is the broad goal worth emulating in policies for

commercializing government managed technologies.

4.3 Challenges for the near future

The diffusion of broad-band access, the widely forecast future for this industry, seems to be taking

on a more typical pattern for new technology, where technical and commercial constraints shape the pattern

of diffusion. It is unclear what the lowest cost method for the delivery of broadband services will be. It is

also unclear what type of services will motivate mass adoption of costly high speed access to the home.

There are technical limitations to retrofitting old cable systems and with developing DSL technology over long

distances. It is unclear how many people will be willing to pay for such high-speed services. These

uncertainties cloud all forecasts. However, unlike the past, there will not be two decades of incubation of

broadband technology by only government sponsored researchers. Hence, there is no reason to anticipate

anything like the speed of diffusion found in the dial-up market, nor take for granted that ubiquity will arise as

easily (for more, see, e.g., Weinberg, 1999, or Werbach, 1997).

This observation would seem, at first blush, to suggest that this history sheds little light on the future

-- that past and future challenges are too unique to their time for comparison. However, that conclusion is a

35

bit hasty. Looking forward in the ISP industry, it is possible to identify some technical, commercial and

structural challenges which resemble those of the past and which will alter the contours of behavior and

outcomes. I will discuss some of these, cognizant that restructuring is still taking place and changing

sufficiently fast, so that any discussion runs the risk of becoming obsolete as soon as it is written.

Lesson 1: The past does offer guidance for understanding patterns of restructuring. The

names of the firms may change and so too may the specifics of the strategies, but the absence of uniformity

in the development of Internet access business models should persist into the future. New applications for

Web technology are still under development because the technology has potential beyond its present

implementations. Not all local markets will experience the same type of competitive choices in access, nor

should they. Not all vendors will see the same opportunities and these differences arise for sound economic

reasons. Users with more experience still adopt applications closer to the frontier, while users with less

experience still demand more refined applications. Web technology enables these differences to manifest in

new directions and it is not obvious which implementation will succeed with either type of user. In other

words, most of the economic fundamentals leading to structure challenges have not disappeared; hence,

experimentation with new business models will probably continue.

Lesson 2: The subscription model of Internet access will continue to change. Commercial

markets inherited an organizational form from their academic ancestors, modifying it slightly for initial use.

There is no reason to presume that it will maintain the same operational structure under competitive pressure.

Indeed, it is presently under competition from a variety of alternate business models which use dial-up access

to subsidize another activity. There are already hints of these potential changes as some ISPs charge very

little for access and make up for the lost revenue with other services, such as networking, hosting or web

design. AOL has successfully combined access with its "walled garden" of content and AT&T appears intent

on pursuing a unique approach to combining content and access. Other recent innovative firms include

Netzero.com, which is the most successful to date of many firms who have tried to provide access for free

36

and garner revenue through sales of advertising. There are also many other such experiments altering the

explicit definition of basic service, embedding it with more than e-mail, but also with games, streaming,

linking and so on, which has the effect of changing the pricing structure too. It is not crazy to predict that

access, by itself, could become absorbed into a bundle of many other complementary commercial services,

slowly fading as a stand-alone service, as it existed in the academic domain.

Lesson 3: The economics of Internet diffusion lie behind much of the digital divide. Internet

access diffused more easily to some users and in some locations. The margin between adoption and non-

adoption has become popularly known as the "digital divide." If some of these outcomes are understood as

temporary results of a young diffusion process, then many of the differences between those with virtual

experience and those without can be framed as the by-product of the economic factors shaping this diffusion

episode. Within business the important factors influencing adoption are the density of the location of the

business, the availability of basic computer support services nearby, and a firm's previous investment in IT.

At the home the important determinants are availability (which is influenced by density) as well as the same

factors behind the diffusion of PCs: age, education and income especially, and also race for some income

levels. It follows, therefore, that policies aimed at digital divide issues, such as the E-rate program, should not

address those factors which are only temporary and will resolve themselves through market forces without

government intervention. Instead, government programs should target factors which are likely to be more

durable over time and which lead to division in adoption behavior; such as density of location, income,

education and race.

Lesson 4: Geographic pervasiveness introduces new economic considerations. There is one

additional reason to expect the typical business model to remain unsettled. Geographic pervasiveness has

entered into calculations today and it was not a relevant consideration at the outset of commercialization. The

pervasiveness of the Internet across the country (and the developed world) changes the economic incentives

to build applications on top of the backbone, and alters the learning process associated with its commercial

37

development. All ISPs now depend on each other on a daily level in terms of their network security, reliability

and some dimensions of performance. Many new applications -- e.g., virtual private networking, voice

telephony over long distances, multi-user conferencing, some forms of instant messaging, and gaming --

require coordinating quality of services across providers. It is still unclear whether new business models are

needed to take advantage of applications which presume geographic pervasiveness. If so, it will provide a

commercial advantage to the firms with national backbones and assets. Pervasiveness also changes the

activities below the backbone in the vertical chain. It has altered the scale of the market for supplying goods

and services to the access industry, altering the incentives of upstream suppliers, equipment manufacturers or

middle-ware software providers, to bring out new services and inventive designs for the entire network. This

factor was also not present in the academic network and it is unclear how it will influence the structure of the

industry moving forward.

Lesson 5: Is there a need for new communications policy for the new millennium? Until

recently, the pace of technical change in most communications services was presumed to be slow and easily

monitored from centralized administrative agencies at the state and federal level. It is well known that such a

presumption is dated, but it is unclear what conceptual paradigm should replace it. This paper illustrated how

vexing the scope of the problem will be. In this instance, ISPs addressed a variety of commercial and

structural challenges with little government interference, but under considerable technical and commercial

uncertainty. This occurred because many legacy regulatory decisions had previously specified how

commercial firms transact with the regulated public switch network. These legacy institutions acted in

society's interest in this instance, fostering experimentation in technically intensive activities, enabling

decentralized decision-making to shape commercial restructuring in specific place and time periods. To put it

simply, it was in society's interest to enhance the variety of approaches to new commercial opportunities and

existing set of regulations did just that. However, going forward it is unclear whether these legacy institutions

are still appropriate for other basic staples of communications policies, such as whether a merger is in the

38

public interest, whether incumbent cable firms should be mandated to provide open access, whether

communications infrastructure should be subsidized in under-served areas, and whether Internet services

should be classified as a special exemption, immune from taxation and other fiscal expenses. Hence, this

industry is entering an era where market events and unceasing restructuring will place considerable tension on

long-standing legal foundations and slow regulatory rule making procedures.

39

References

Augereau, Angelique, and Shane Greenstein [1999]. "The Need for Speed in Emerging CommunicationsMarkets: Upgrades to Advanced Technology at Internet Service Providers."http://www.kellogg.nwu.edu/faculty/greenstein/images/research.html

Boardwatch Magazine, Directory of Internet Service Providers, Littleton, CO.

Bresnahan, Timothy, and Shane Greenstein [1999]. "Technological Competition and the Structure of theComputing Industry," Journal of Industrial Economics, 47, 1-40.

Bresnahan, Timothy, and Manuel Trajtenberg [1995]. "General Purpose Technologies: Engines of Growth?"Journal of Econometrics, 65, 83-108.

Clemente, Peter C. [1998]. The State of the Net: The New Frontier, New York: McGraw-Hill.

Cohen, Linda, and Roger Noll [1990]. The Technology Pork Barrel, Brookings.

Cowan, Robin [1988]. Nuclear Power Reactors: A Study in Technological Lock-in, Working Paper 88-33,C.V. Starr Center for Applied Economics, New York University.

Cusumano, Michael and David Yoffie [1998]. Competing on Internet Time, Lessons from Netscape and ItBattle With Microsoft, New York: The Free Press.

Demsetz, Harold [1988]. “The Theory of the Firm Revisited,” Journal of Law, Economics, andOrganization, 4, 159-178.

Downes, Tom, and Greenstein, Shane [1998]. "Do Commercial ISPs Provide Universal Access?" in (Ed)Sharon Gillett and Ingo Vogelsang, Competition, Regulation and Convergence: Selected Papers from the1998 Telecommunications Policy Research Conference, Lawrence Erlbaum Associates.

Flamm, Kenneth [1989]. Targeting the Computer, Government Support and International Competition,Brookings Institute, Washington, D.C.

Greenstein, Shane [1999]. "Building and Developing the Virtual World: The Commercial Internet AccessMarket," http://skew2.kellogg.nwu.edu/~greenste/research.html

Griliches, Zvi [1957]. "Hybrid corn: An exploration in the economics of technological change," Econometrica,25, 501-522.

Helpman, Elhanan [1998]. General Purpose Technologies and Economic Growth, The MIT Press.

Kalakota, Ravi and Andrew Whinston [1996]. Frontiers of Electronic Commerce, Addison-Wesley, Reading,MA.

Katz, Barbara, and Almiran Phillips [1982]. "The Computer Industry," in Richard Nelson R (Ed.) Governmentand Technical Progress: A Cross-Industry Analysis, New York: Pergamon Press, pp. 162-232.

40

Kolstad, Rob [1998]. "Becoming an ISP," www.bsdi.com. January.

Maloff Group International, Inc.[1997]. "1996-1997 Internet Ac cess Providers MarketplaceAnalysis," Dexter, MO, October.

Mowery, David and Nathan Rosenberg [1992]. Technology and the Pursuit of Economic Growth, CambridgeUniversity Press.

National Telecommunication and Information Administration [1999]. Falling Through the Net: Defining theDigital Divide. US Department of Commerce, July.

Nelson, Richard, and Sidney Winter [1977]. An Evolutionary Theory of Economic Change, Cambridge, MA:Belknap Press.

Nie, Norman and Ebring, Lutz [2000]. Internet and Society: A Preliminary Report, Stanford Institute for theQuantitative Study of Society, Stanford University.

Sidek, Greg and Dan Spulber [1998]. “Cyberjam: The Law and Economics of Internet Congestion of theHarvard Journal of Law and Public Policy, 21, 2.

Strover, Sharon [1999]. "Rural Internet Connectivity." Working Paper, University of Texas at Austin.

Ventresca, Mark, Rodney Lacey, Michael Lounsbury, and Dara Szyliowicz [1998]. "Industries asOrganizational Fields: Infrastructure and Formative Dynamics in U.S. Online Database Services." A WorkingPaper in the "Business and Government" Series of the Institute for Policy Research, Northwestern University.

Waldrop, Mitchell [2001, forthcoming], Technology of Enchantment. Viking Press.

Weinberg, Jonathan [1999]. "The Internet and Telecommunications Services, Access Charges, UniversalService Mechanisms, and Other Flotsam of the Regulatory System." Yale Journal of Regulation, Spring.

Werbach, Kevin [1997]. A Digital Tornado: The Internet and Telecommunications Policy. FCC, Office ofPlanning and Policy Working Paper 29, March.